Automatic Fault Detection for Marine Diesel Engine Degradation in Autonomous Ferry Crossing Operation

Abstract

The maritime industry generally anticipates having semi-autonomous ferries in commercial use on the west coast of Norway by the end of this decade. In order to schedule maintenance operations of critical components in a secure and costeffective manner, a reliable prognostics and health management system is essential during autonomous operations. Any remaining useful life prediction obtained from such system should depend on an automatic fault detection algorithm. In this study, an unsupervised reconstruction-based fault detection algorithm is used to predict faults automatically in a simulated autonomous ferry crossing operation. The benefits of the algorithm are confirmed on data sets of real-operational data from a marine diesel engine collected from a hybrid power lab. During the ferry crossing operation, the engine is subjected to drastic changes in operational loads. This increases the difficulty of the algorithm to detect faults with high accuracy. Thus, to support the algorithm, three different feature selection processes on the input data is compared. The results suggest that the algorithm achieves the highest prediction accuracy when the input data is subjected to feature selection based on sensitivity analysis.

Online Fault Detection in Autonomous Ferries: Using Fault-type Independent Spectral Anomaly Detection

“…A third fault type is used for the final test of the proposed algorithm: a malfunction of the frequency-operated fan controlling the secondary cooling system in the engine. This fault, which appears in our previous work [14], is hereinafter referred to as the cooling fault. One normal operation data set, one turbo degradation data set, and one air filter degradation data set are collected from each profile.…”

“…The sensors installed on autonomous ferries can be utilized to accumulate and collect normal operation data to use a semisupervised learning framework. A VAE was used for anomaly detection in [9] and [14]. In both studies, the maximum acceleration in faulty degradation data was estimated and used as the fault detector.…”

“…where n is the number of input features andŷ i and y i are the i th reconstructed and target measurement, respectively. As in [9] and [14], the VAE is structured with two hidden layers (h 1 , h 2 ) in the encoder, z units in the latent layer, and two hidden layers (h 2 , h 1 ) in the decoder. However, the number of hidden units in each layer differs from the previous studies as they are determined related to the number of input features n…”

“…During the data collection process, the engine is driven by two different engine load profiles. Similar to [14], the two engine load profiles aim to replicate real-life autonomous ferry crossings on the west coast of Norway. First, the ferry is off-loading and on-loading vehicles before it leaves shore at a safe and constant velocity.…”

“…The variational autoencoder (VAE) is the selected DNN as it outperforms a feedforward neural network (FNN) with one hidden layer, the traditional autoencoder (AE), and the long short-term memory (LSTM), in terms of reconstruction-based fault detection for maritime components in [9]. Similar to [14], a replicated autonomous ferry crossing operation is used to produce two engine load profiles. These profiles reflect different environmental conditions affecting the ferry.…”